Delaying single analog pulse

Thread Starter

JPilot

Joined Nov 22, 2022
2
Hello all,

As title says I would like to delay an analog pulse (something like this)
1669167347462.png
by a particular amount of time (say 5 usec) using hardware. These pulses are random (not periodic), occurring only ~once/sec, pulse width ~400 ns, amplitude ~10 mV. It doesn't matter if the delay distorts the waveform a bit, as all I need to know is if the pulse is above a given threshold amplitude.

Also I need to do this for 64 separate signals (from SiPMs if that helps) so I'm looking for a solution which minimizes the cost/complexity in terms of components/amplification. Ideally I would like to find an off the shelf IC that does this, but everything I can find only takes TTL inputs and max delays of ~500 ns. Does anyone have any ideas?

Cheers!
 

camerart

Joined Feb 25, 2013
3,291
Hello all,

As title says I would like to delay an analog pulse (something like this)
View attachment 281242
by a particular amount of time (say 5 usec) using hardware. These pulses are random (not periodic), occurring only ~once/sec, pulse width ~400 ns, amplitude ~10 mV. It doesn't matter if the delay distorts the waveform a bit, as all I need to know is if the pulse is above a given threshold amplitude.

Also I need to do this for 64 separate signals (from SiPMs if that helps) so I'm looking for a solution which minimizes the cost/complexity in terms of components/amplification. Ideally I would like to find an off the shelf IC that does this, but everything I can find only takes TTL inputs and max delays of ~500 ns. Does anyone have any ideas?

Cheers!
Hi,
I think you could use PICs, which have analogue inputs, and a delay can be programmed.
I don't use Arduinos, but I'm sure they would work too.
Cheers, C
 

WBahn

Joined Mar 31, 2012
27,903
You picture shows a current pulse, but you talk about a 10 mV amplitude.

Assuming you mean 10 mV, that might be pretty hard to work with unless you amplify it.

If all you need to know is whether it is above some threshold, why do you need to delay it? Can't you just determine if it is above the threshold when it occurs and then delay whatever digital signal results however long you need to? That you can do either with a MCU or with a digital delay line, which you can get that have programmable delays from tens of nanoseconds range to tens of seconds range.

One potential option is a bucket-brigade device (analog delay line). They used to be everywhere, but now they are hard to come by (new, anyway). I think they are still made for guitar pedals or something similar.
 

ronsimpson

Joined Oct 7, 2019
2,540
The pulse is about 400nS long and needs to be delayed 5uS. The entire pulse needs to be saved, delayed, and then output.
If we use a "computer" like mentioned in #2, the Arduino cannot do one instruction in 5uS and most PICs can only do a couple instructions in 5uS. I don't know how accurately you need the output to be. If the 400nS pulse was divided into 100 pieces and put into memory it could be delayed any amount of time. This would require reading in data every 4nS and send data out every 4nS. This is never happening in a PIC. Using this method, you need the front end of a 100mhz digital oscilloscope.

On the other hand, if you don't care what the pulse looks like, this is easy. Delay=5uS, pulse width=400nS. This is very easy if a pulse only comes less than one/5uS. (only one pulse at a time) Please comment on this and one of us will make a delay circuit.
1669212999649.png
-----------edited-------------
My first idea has a fixed 400nS output but now I see how to make the output pulse width match the input pulse width is that is important. Very simple with one IC. How accurate must the delay be. 5%??
74LS123
 
Last edited:

DickCappels

Joined Aug 21, 2008
9,513

MisterBill2

Joined Jan 23, 2018
13,805
The delay line concept described by DC can work, you can also use a length of coaxial cable as a delay line, which should be simple to make many of them the same. But probably you would also need an amplifier to drive the line. A video amp such as the LM733, or equal, would work. If the line loss is not excessive at the required delay you might get away with simply reducing the trigger level on the detector.
 

WBahn

Joined Mar 31, 2012
27,903
It would take something like a kilometer of coax to get 5 µs of delay. I'm guessing that making 64 kilometer long cables all the same length would be a bit easier said than done and that it would likely not qualify for the goal of minimizing the cost.
 

camerart

Joined Feb 25, 2013
3,291
It would take something like a kilometer of coax to get 5 µs of delay. I'm guessing that making 64 kilometer long cables all the same length would be a bit easier said than done and that it would likely not qualify for the goal of minimizing the cost.
Hi W,
Would that be the same for a radio signa;l?
C
 

WBahn

Joined Mar 31, 2012
27,903
Hi W,
Would that be the same for a radio signa;l?
C
The delay is nothing more than how long it takes the signal to propagate from one end of the cable to the other at the speed of the waveform in that medium. For a lot of conductor materials and transmission lines, as well as optical fiber, the speed of propagation is approximately 2/3 of the speed of light in a vacuum, or about 200 m/µs. There are cases where it is quite a bit lower or quite a bit higher, but this is a pretty useful general rule-of-thumb. There is always some wavelength dependence on that as well, but not enough to matter for this purpose.
 

Thread Starter

JPilot

Joined Nov 22, 2022
2
Thanks for the responses everyone, let me give a bit more info on my application. What I'm trying to do here is identify pulses from 8 different SiPMs (type of photodetector) with only 1 scope/digitizer channel. So my plan was to delay the first SiPM signal by 1 usec, the second signal by 2 usec, etc. Then I can combine all those signals into the same channel (superimposing them) and identify which SiPMs pulsed based on time elapsed between a given pulse and some initial trigger. This is why I'm trying to do this analog rather than digital.

My first idea has a fixed 400nS output but now I see how to make the output pulse width match the input pulse width is that is important. Very simple with one IC. How accurate must the delay be. 5%??
74LS123
5% should be fine I think, because I would still be sampling near enough the peak to see it.

Since you don't need great waveform fidelity, you might want to look at a lumped constant delay line approximation.
View attachment 281269

Like a transmission line, it needs to have its output and preferably input terminated in the line's characteristic impedance. Low fidelity means you can get away with fewer stages.
This web page: https://electronics.stackexchange.com/questions/377176/lumped-element-delay-lines-using-lc-circuit
gives you the formulas.
So t=sqrt(LC), and => a single stage with 1 uH inductor and 1 uF capacitor would get me a 1usec delay? This seems very easy, am I missing some catch?

You picture shows a current pulse, but you talk about a 10 mV amplitude.

Assuming you mean 10 mV, that might be pretty hard to work with unless you amplify it.

If all you need to know is whether it is above some threshold, why do you need to delay it? Can't you just determine if it is above the threshold when it occurs and then delay whatever digital signal results however long you need to? That you can do either with a MCU or with a digital delay line, which you can get that have programmable delays from tens of nanoseconds range to tens of seconds range.

One potential option is a bucket-brigade device (analog delay line). They used to be everywhere, but now they are hard to come by (new, anyway). I think they are still made for guitar pedals or something similar.
The SiPM outputs a current pulse but I am reading it across a resistor as a voltage pulse. To do this digitally, could I put the signals into comparators to convert to TTL, then use digital delay line ICs like this?

https://www.mouser.com/datasheet/2/256/DS1110L-DS1110LE-500-1513729.pdf

Thanks all!
 

DickCappels

Joined Aug 21, 2008
9,513
So t=sqrt(LC), and => a single stage with 1 uH inductor and 1 uF capacitor would get me a 1usec delay? This seems very easy, am I missing some catch?
That's about it. Note that the delay line is also a low pass filter, so your pulse will be smeared out in time. To reduce this effect you can use multiple LC sections, each of which has a wider bandwidth and less delay.
 

MisterBill2

Joined Jan 23, 2018
13,805
AS there are evidently 64 (8x8) channels from a silicon photo multiplier, an area to investigate could be the array processing scheme used in ultrasound medical scanners. They already have that scheme working very well. So there is probably that hardware available in a compact package.
Sometimes it works to copy what already has been a success.
 
Top